Thanks for sharing this document ( http://goo.gl/dxRdGS ).
The document does not seem to detail what specific problems are being worked around on "problematic graphics cards". It only mentions taking "corrective action".
What I would really like to understand --- and that is what would affect the most my position on these questions --- is whether there exist broad categories of problems that can be reasonably well detected from the GL strings, and not from other existing WebGL getters.
The document asserts that the Google Maps team has identified such broad issues, and has successfully worked around them by using GL strings. But that information alone is hardly actionable for me, not knowing the underlying data.
I do understand that it would be difficult to share publicly data that would describe some GPUs as "poorly performing". But perhaps you could at least share the data of what categories of issues are being worked around by using GL strings (e.g. "branching in fragment shaders is slow") without embarrassing any particular GPU vendor?
On 14-01-13 06:39 PM, Jennifer Maurer wrote: